@nir_moneta

Postdoc in Summerfield Lab, Oxford | 🏳️‍🌈 | currently into task representations in 🧠 and 🤖 (NaN/NaN) https://xitter.cf/t.co/AWY71qDNGa

 Berlin, Germany
Joined November 2017
  • Tweets139
  • Following735
  • Followers448
13Media
What happens when choices can lead to diff outcomes, depending on context? See our work on how state representations in vmPFC mediate relevant AND irrelevant expected values signals, out in @NatureComms w/ @mona_garvert @haukeren & @nico_schuck [1/n] rdcu.be/ddq89
2
16
4
74
12,507
thanks @NeuroCellPress for publishing our paper on the "recipe for intelligence" (with @neuro_kim). Now out cell.com/neuron/fulltext/S08…

The recipe for intelligence in natural and artificial systems

This review by Summerfield and Stachenfeld surveys how the dialog between neuroscience and AI research has evolved over the past decade. They highlight continued opportunities for synergy between...

cell.com
17
51
4,903
Nir Moneta retweeted
LJDM@LJDM1Mar 11
Happening today! Join us for Dr Moneta's talk — in person in Room 313 (UCL EP, 26 Bedford Way), or online. All are welcome! P.S. I have been told the talk should also be accessible to a non-neuroscience audience!
LJDM@LJDM1Mar 5
Next Wednesday (March 11th) we are turning more neuroscience-y with Dr @nir_moneta joining us from @UniofOxford to talk about how reward learning influences non-spatial cognitive maps. Please join us in person (full details: surl.li/mqtofs). All are welcome!
1
2
177
This Wednesday at UCL, come by if you’re around!
LJDM@LJDM1Mar 5
Next Wednesday (March 11th) we are turning more neuroscience-y with Dr @nir_moneta joining us from @UniofOxford to talk about how reward learning influences non-spatial cognitive maps. Please join us in person (full details: surl.li/mqtofs). All are welcome!
2
102
Happy to share that after successfully defending my PhD last April with the great @nico_schuck , I will start a Postdoc with @summerfieldlab on 1st of September. Grateful for the past, excited for the future!
25
694
Attending #CCN2025?  Come check out our poster tomorrow (Tuesday, A90) on how reward morphs neural representations of non-spatial cognitive maps in the hippocampus. Work w. @lui_anta @NitschAlex @TheCharleyWu @doellerlab @nico_schuck 2025.ccneuro.org/poster/?id=…
6
354
Are you coming to @RLDMDublin2025 #RLDM2025? Join our workshop on Gamification in Computational and Cognitive Neuroscience with an amazing speaker lineup! 🎮🧠🪄 w/ @OndrejZika @ingrdmrtn
3
1
6
769
It’s a Preprint! 👋 We show how we can dissociate perceptual from value-based mechanisms of generalisation + that stronger generalisation in anxiety is associated with value rather than perception. w/ @OndrejZika @nico_schuck @BernhardSpitzer osf.io/preprints/psyarxiv/yh… 1/n
2
5
1
12
1,773
🚨Very excited that our (@lui_anta @ingrdmrtn & myself) RLDM 2025 Workshop "Game on: Leveraging Gamification to Understand Cognitive and Computational Mechanisms" has been accepted! We will soon be opening a call for contributed talks - stay tuned if you are interested!
3
18
705
Nir Moneta retweeted
Yael Niv@yael_niv20 Dec 2024
Friends, this is urgent. Families in Gaza are starving and need our help. I have donated to and been following this soup kitchen since they started. They provide hot meals to thousands & they need funds to keep going. Please consider donating & sharing. gofund.me/aab9459d

Donate to Hot meals 4 Starved Palestinian Kids in north Gaza, organized by Hani Almadhoun

Do you know what it’s like to watch your family starve? I do. My… Hani Almadhoun needs your support for Hot meals 4 Starved Palestinian Kids in north Gaza

gofundme.com
5
11
34
3,055
Super excited to be going to #NeurIPS to present new work on softly state-invariant world models! We introduce an info bottleneck making world models represent action effects more consistently in latent space, improving prediction and planning! Reach out if you want to meet!
1
27
1
100
16,278
Excited to share our review, w. @ShanyGrossman (shared 1st!) and @nico_schuck where we discuss representational spaces in OFC/vmPFC and in deep RL models, featuring values, states and much more! cell.com/trends/neuroscience…. Fresh from the (virtual) print now in @TrendsNeuro , 1/4

Representational spaces in orbitofrontal and ventromedial prefrontal cortex: task states, values,...

The orbitofrontal cortex (OFC) and ventromedial-prefrontal cortex (vmPFC) play a key role in decision-making and encode task states in addition to expected value. We review evidence suggesting a...

cell.com
1
22
1
75
8,398
more replies
3/4 We discuss our previous work where we showed states are decodable from vmPFC, alongside values, and modulate irrelevant values' influence on behavior. With other studies, this supports complex rep in vmPFC/OFC to allow flexible goal-directed behavior. nature.com/articles/s41467-0…

Task state representations in vmPFC mediate relevant and irrelevant value signals and their...

Nature Communications - The ventromedial prefrontal-cortex (vmPFC) encodes expected value signals that contribute to choices. Here the authors show that during decision-making the human...

nature.com
1
1
219
4/4 Lastly, we discuss the potential contribution of late layers in deep RL models to understanding emergent reps in OFC/vmPFC, which go beyond value, and raise outstanding questions for future research.
145
Excited to announce that this work has now been published in @PNASNews pnas.org/doi/10.1073/pnas.24… We propose a computational account of how humans decompose complex dynamical experiences into subprocesses and reuse these in an abstract, compositional manner in novel situations.

Decomposing dynamical subprocesses for compositional generalization | PNAS

A striking feature of human cognition is an exceptional ability to rapidly adapt to novel situations. It is proposed this relies on abstracting and...

pnas.org
Lennart Luettgau@LLuettgau3 Sep 2023
As promised, here’s the summary thread🧵: The study presented in this preprint investigates how humans decompose dynamical subprocesses and leverage the abstracted subprocesses for compositional reuse of experience in new situations. 1/n
3
13
1
50
7,534
This is finally out today in @PLOSCompBiol - find the final updated and improved version here: journals.plos.org/ploscompbi… with @nico_schuck @summerfieldlab @SaxeLab @PMuhleKarbe and twitterless Léo

Abrupt and spontaneous strategy switches emerge in simple regularised neural networks

Author summary Insights, or aha-moments, are a remarkable phenomenon in human cognition that is unique in a number of ways: they are accompanied by a powerful subjective experience, occur abruptly...

journals.plos.org
Anika Löwe@anika_loewe23 Feb 2023
New Preprint! We often learn gradually, but sometimes things occur to us in a flash, aka aha-moment. Why? We explored the comp. mechanisms of spontaneous periods of fast learning & show they arise even in simple NNs. arxiv.org/abs/2302.11351
1
13
2
44
7,490
Interested in getting hands-on skills in cognitive neuroscience? We are looking for 1-2 research assistants to support our research at @PB_UniHH @unihh Come work with our amazing team. Hamburg students from all departments are eligible. More about us: schucklab.gitlab.io
11
8
1,831
🚨 New preprint on the role of eating disorder symptoms in accelerating response vigor during foraging decisions 💫 w/ Sam Hall-McMaster 📜Link: osf.io/preprints/psyarxiv/kz… 💻Code and data: github.com/ozika/ed-foraging… Funded by @AvHStiftung @mpib_berlin
1
6
14
1,079
Can LLMs do reinforcement learning in-context - and if so, how do they do it? Using Sparse Autoencoders, we find that Llama 3 relies on representations resembling TD errors, Q-values and even the SR to learn in three RL tasks in-context! Co-lead with the inimitable @can_demircann
3
35
10
201
26,476
Excited to share our brand-new work shedding some light on the neural mechanisms behind one of human’s coolest cognitive feats: compositional generalization of structural knowledge! A Tweeprint-Thread 🧵 1/n
1
7
3
27
4,064
Load more